迁移增量启发式动态规划及污水处理应用  

Transferable Incremental Heuristic Dynamic Programming With Wastewater Treatment Applications

作　　者：王鼎 李鑫 WANG Ding;LI Xin(Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China;Beijing Key Laboratory of Computational Intelligence and Intelligent System,Beijing 100124,China;Beijing Laboratory of Smart Environmental Protection,Beijing 100124,China;Beijing Institute of Artificial Intelligence,Beijing 100124,China)

机构地区：[1]北京工业大学信息学部,北京100124 [2]计算智能与智能系统北京市重点实验室,北京100124 [3]智慧环保北京实验室,北京100124 [4]北京人工智能研究院,北京100124

出　　处：《北京工业大学学报》2025年第3期277-283,共7页Journal of Beijing University of Technology

基　　金：科技创新2030重大项目(2021ZD0112302,2021ZD0112301);国家自然科学基金资助项目(62222301,61890930-5,62021003);北京市自然科学基金资助项目(JQ19013)。

摘　　要：针对污水处理系统中的溶解氧(dissolved oxygen,DO)质量浓度控制问题,提出一种迁移增量启发式动态规划(transferable incremental heuristic dynamic programming,TI-HDP)算法。针对污水处理过程的特性,该算法通过将控制变量的更新方式改进为增量形式,提升了算法的抗干扰能力,并弱化了与增量式比例-积分-微分(proportional-integral-derivative,PID)算法之间的结构差异。基于数据驱动的思想,通过利用PID算法所产生的历史数据,成功地将传统控制领域中的专家经验迁移到TI-HDP算法框架中,保证了TI-HDP算法前期控制策略的稳定性。仿真结果表明:与PID算法和传统的启发式动态规划算法相比,所提算法对DO质量浓度具有更高的控制精度。A transferable incremental heuristic dynamic programming(TI-HDP)algorithm is proposed for the control problem of the dissolved oxygen(DO)mass concentration in the wastewater treatment system.Considering the characteristics of the wastewater treatment process,this algorithm improved the anti-interference ability and weakened the structural disparity with the incremental proportional-integral-derivative(PID)algorithm by improving the updating method of the control variable into the incremental form.Based on the data-driven idea and by utilizing the historical data generated under the action of the PID algorithm,the expert experience in the traditional control field was successfully integrated into the framework of the TI-HDP algorithm,which ensured the stability of the control strategy of the TI-HDP algorithm in the early stage.Simulation results show that the TI-HDP algorithm achieves a higher control accuracy for the DO mass concentration than the PID algorithm and the traditional heuristic dynamic programuing(HDP)algorithm.

关 键 词：启发式动态规划(heuristic dynamic programming HDP) 智能控制 知识迁移 非线性系统 神经网络 污水处理 

分 类 号：U461[机械工程—车辆工程] TP308[交通运输工程—载运工具运用工程]